Robots are used in many precision manufacturing processes. For example, robots are used to precisely align lenses before digital camera sensors, such as in the manufacture of cell phones and backup cameras for automobiles. In other examples, robots align ends of optical fibers before lasers or light sensors in the manufacture of telecommunication and computer network equipment. Many of the lenses are quite small, on the order of several millimeters in diameter, and must, therefore, be positioned with high precision, often on the order of about ±5 μm or less, relative to the sensors or lasers.
To keep costs down, less-than-precise methods are often used to manufacture optical elements for the lenses and to mount the optical elements in lens housings. Consequently, the optical elements and lenses are often not uniform, from piece to piece. That is, dimensions and symmetries of the components often vary from lens to lens or other optical components, resulting in variations in focal length and orientation of the optical axes of the components.
To compensate for such variations, several known methods are used to custom align each lens to its corresponding image sensor. One method involves mounting a finely threaded bracket to the sensor and mounting a group of optical elements in a complementarily threaded barrel. The distance between the optical elements and the sensor can then be adjusted as the lens barrel is threaded into the bracket. Once the optical elements are disposed a desired distance from the sensor, the barrel may be fixed, relative to the bracket, to prevent further rotation. Unfortunately, this method allows adjustment of only the distance between the optical elements and the sensor along the z axis (the optical axis). Thus, this method is referred to as aligning with only one degree of freedom. Such an alignment methodology cannot compensate for all potential asymmetries in the optical elements or imperfect alignment of the optical elements within the barrel.
A more sophisticated method, developed by Automation Engineering, Inc. (“AEi”), Wilmington, Mass., involves attaching the lens barrel (which in this case does not require threads) to the image sensor or its substrate or a camera housing that includes the image sensor by an adhesive with camera module or other opto-electronic product component geometries such that more than one degree of freedom alignment is possible. The position of the lens barrel, relative to the image sensor, is adjusted in free space by a robot, which then holds the lens barrel in position until the adhesive cures sufficiently to prevent objectionable creep once the robot releases the lens barrel. Using this method, the position of the lens may be adjusted along all three linear axes (x, y and z), and pitch (rotation about the x axis) and roll (rotation about the y axis) of the lens may be adjusted, to achieve a lens position and orientation, relative to the sensor, that compensates for imperfections in the optical elements and in the way the optical elements are mounted within the barrel with respect to the position of the image sensor. This method is referred to as aligning about five degrees of freedom.
A refinement of this method, also develop by Automation Engineering, Inc., involves the robot rotating the lens about the optical axis of the lens or about the z axis to optimize image quality, to compensate for angular lens asymmetries in this axis. Note that alignment about the optical or z axis is generally not possible with the threaded barrel alignment method, because it is highly unlikely to simultaneously achieve both a desired lens-to-sensor spacing (by threading the barrel in or out) and a desired lens rotation angle (by rotating the barrel). Adding this refinement to the 5-degree of freedom alignment method provides a 6-degree of freedom alignment.
“Passive alignment” involves aligning components to each other based on design specifications of the components, using precision mechanical fixtures, tooling, physical characteristics (fiducials) on surfaces of components and the like. For example, a target design focal length of a lens may be used to position the lens, relative to a sensor. However, passive alignment assumes components perfectly conform to their theoretical design specifications. This, of course, does not typically occur with real-world products. Thus, passive alignment methods are typically unable to compensate adequately for piece-to-piece variations in components, such as lenses, unless each piece is individually tested to ascertain its actual specifications.
In contrast, “active alignment” involves measuring one or more key performance attributes of a combination of components during manufacture and using the measured attributes in a feedback loop to control the manufacture. For example, a visual target, such as a test pattern, may be disposed within a viewing angle of a lens-and-image-sensor combination while a robot steps the lens through a series of discrete positions (ex., along the x, y and z axes) and orientations (ex., tips, tilts and rotations about the z axis). A computer analyzes image data from the sensor at each step and, based on this analysis, the computer controls the robot to position and orient the lens for optimum image quality. The lens is then fixed in position, relative to the sensor, such as by an adhesive. Active alignment is, therefore, able to compensate for piece to piece variations in components.
While active alignment requires hardware and software not necessary for passive alignment, active alignment may not necessarily be more expensive than passive alignment. For example, active alignment can often be performed using less accurate and, therefore, less expensive fixturing and motion stages in a given manufacturing station. Active alignment can also provide higher manufacturing yield, which results in reduced manufacturing costs. However, stepping through all the test positions and orientations takes time. Manufacturers constantly seek to shorten cycle times and, thus, speed up manufacturing processes.